great – aal-2016-023 · great – aal-2016-023 34 4 scent module compared to the field trial...

GREAT – AAL-2016-023

31

Figure 37: 3D model of Raspberry PI + Pi-DAC Zero holder for mounting inside Z150 speaker

The Logitech amplifier board is connected to the Pi-DAC Zero via the 2x4 header on the

Pi-DAC Zero, which provides line-out left/right, as well as GND and a 5V power supply (see

Figure 38 for details of the wiring). Note that the original power supply connection is

unmounted from the speaker system.

Figure 38: Wiring of the Pi-DAC 2x4 header to the Logitech board inside the speaker

As an alternative to building the sound module hardware into the Z150 speakers, a

separate housing has been designed, in case the Raspberry Pi and Pi-DAC Zero should be

connected to an existing sound system or alternative speakers. Figure 39 shows the 3D

model of the case divided into two parts.


32

Figure 39: 3D model of separate housing for sound module logics

During the field test, the Logitech speakers with the built-in sound module have been used

(see Table 3 for a list of required components).

Table 3: GREAT Sound Module hardware component list

1 Raspberry Pi Zero W board

1 IQaudIO PiZero DAC

1 Micro USB Power Supply, 1A

1 microSDHC card, 16GB, Class 10

(industrial)

1 Logitech Z150 Speaker Pair

The external sound module has been used in combination with a high-quality active

speaker (Pioneer RM05) for testing inside the GREAT cabin. This speaker has an advertised


33

frequency range of 40-50.000 Hz, and therefore a significantly better reproduction of ultra-

sonic sound range.

Figure 40: External sound module with a Pioneer RM05 Speaker featuring a frequency range of 40-

50.000Hz

3.3 Speaker Covers

To prevent accidental adjustment of the speaker volume by the built-in volume

adjustment knobs, a selection of speaker covers has been designed. For the field tests, the

minimalist grey cover has been preferred (see Figure 41).

Figure 41: Speaker covers to prevent accidental adjustment of volume level


34

4 Scent Module

Compared to the field trial phase the scent module was reworked significantly: the casing

presented in D2.1 as an outlook was further adapted, one-part 3D-printed and the other

vacuum-formed. The pumping mechanism was changed from piston-based to spring-

loaded. Further smaller changes will be presented in the sub-chapters below.

4.1 Basic Considerations

The GREAT-system requires a scent-module, which can adequately enhance the room

atmosphere by dispersing respective odors. This basic requirement has not changed since

the beginning of the project and we continue using spray-based dispersion for our field-

test phase. Due to feedback from the tests so far hardware upgrades on the in- and

outside of the device were necessary. The scent module should look more appealing and

be able to resist harsher external influences (the previously used acrylic glass could shatter

and did not look professional enough).

As stated further below, the pump mechanism from the functional test phase was not

satisfying (a small portion of the droplets were too big and left stains on surrounding

furniture), therefore the activation mechanism of the scent module was also improved.

Changing the method of dispersion completely was considered (to either ultrasonic or

heat vaporization), but next to the downsides already mentioned in D2.1 – involuntary

vaporization, ultrasonic noise (which conflicts with the sound module) and alteration of

the scent molecules (due to heat) – financially it was not an option to develop a

completely new scent module.

4.2 Mechanical Design

Feedback during the early field-test-phase stated that the piston-based pumping

mechanism created droplets of an undesirable size (too big). These droplets did not

vaporize on their way from the bottle to the ground but left visible drops on the furniture or

floor below. After the final vaporization there, a dry stain could still be seen.

We identified the sinusoidal stroke of the piston as the main problem: for a pump-action-

based disperser, the last part of the compression is the most significant one, as the pump-

head is fully loaded with liquid and when pressed appropriately quickly, it will create the

desired scent-cloud. Unfortunately, the sinusoidal stroke does not allow for this, as the

speed of the piston decelerates. Figure 42 shows a sinusoidal stroke, data used for

linearization and a linear approximation, which would continue with constant speed until

the maximum stroke.


35

Figure 42: Sine, linearization and database for linearization and formula for the linear approximation

The error between that approximation and the sine curve can be seen in Figure 43. After

226 degrees, the sine starts to decelerate more and more, which means that during the

last quarter of a stroke, the scent module would dispense at decreasing velocity, which

lead to the undesired droplet size. In figures x and y, a stroke starts at 90 degrees and ends

at 270 degrees. This means, one stroke lasts for 180 degrees (half of a rotation). The

difference between 226 and 270 is 44, which makes up a quarter of one stroke.

Figure 43: Error between sine and linear approximation, absolute [-1, 1] and percentage


36

To avoid this problem, we remodeled the interior parts of the casing, with the major

change being the removal of the piston and the addition of a spring as seen in Figure 44.

The moving parts of the mechanism were previously milled at the FH Vorarlberg. During

the redesign, an iterative design process was used, where each step of the evolution was

3D-printed. Figure 44 shows the four new parts responsible for releasing the scent. The part

seen on top fixes the motor, prevents the gears from slipping away, attaches to one end

of the spring and the lever. As it is the only non-moving part, it will be referenced as

“mount”. The redesigned piston (as seen in the left half of the center row in Figure 44) now

is disconnected from the other parts and only attaches to the linear guide. The final

iterations also contain a screw to stop the piston from falling off the linear guide when

changing the bottles. The eccentric (as seen in the right half of the center row in Figure 44)

is held in place within a gear by a perpendicular screw and catches the lever on each

rotation. The lever (as seen in the bottom row of in Figure 44) is attached to the mount at

two points: a screw at the bottom, allowing it to pivot and with the spring at the top.

Figure 44: Evolution of mount (top), piston (center left), eccentric (center right) and lever (bottom)

The motor and the gears remain, but now they spin an eccentric which in the one half of

a rotation simultaneously stretches a spring and moves the lever away from the bottle,

allowing the pump head to rise to its normal level (seen on the right side of Figure 45). At

the maximum stroke, it moves away from the lever, which is then pulled onto the piston by

the spring, which leads the force to the pump head (seen on the left side of Figure 45).

The motor then continues to spin without load until the eccentric catches the lever again.

Figure 46 shows the mechanic as installed in the field-test prototypes. It was replaced in all

field-test-modules between March and May 2019. For a video of the new functionality,

please visit our twitter page: https://twitter.com/GREAT_AAL/status/1111238884720033792.

The total amount of parts required was reduced and assembling the module is now faster

than before. Replacement parts can be printed at the FH Vorarlberg should any damages

occur. Using our own 3D-printers instead of purchased parts presents a valuable reduction

in costs for our field test prototypes.


37

Figure 45: Left shows the no-load-phase of the motor, right shows the spring-tensioning-phase of the

motor

For the new mechanism we need two additional screws (the ones preventing the piston

from sliding off the linear guide) but less nuts as all but two screws (they can be seen in

Figure 46 on the right, being screwed from the outside into the mount) directly screw into

the 3D-printed material.

Figure 46: Reworked mechanic installed in the rear casing


38

4.3 Casing

The casing consists of two main parts: a front part (evolution as seen in Figure 47), a rear

part, and a small inner casing to protect the microcontroller. In 2018 we also used optional

shields (also seen in Figure 47) to prevent bigger droplets from hitting furniture. However,

these shields have been retired, as the reworked mechanism vaporizes the scent liquid

sufficiently.

Figure 47: Evolution of the 3D-printed front casing (top) and the two variants of shields (bottom)

4.3.1 Front Casing

To create an appealing look, which visually connects light, and scent module, the

previous casing (left casing in Figure 47) was reworked with an external designer (center

casing in Figure 47) who also helped creating the cooling body for the lamp. The final

design (right casing in Figure 47) has rounded edges and (not visible in this picture)

increased space for the motors inside.

The production of the resulting 3D-model was outsourced to a local company, specializing

in small series production. The actual manufacturing-process is vacuum forming, as it

creates a nice-looking finish and has appropriate mechanical durability.

When the Raspberry Pi is connected to its power source, the front casing cannot be

removed, as the plug locks it in place. This was introduced to prevent modules from

spraying at users servicing them. The front casing can also be seen in Figure 48 on the right

side.


39

Figure 48: Left shows the completely mounted rear casing, right shows the front casing

4.3.2 Rear Casing

As before, the rear casing holds all mechanical and electrical parts as well as the bottled

scents (see Figure 48, left side). Previously, it consisted of two identical aluminium bodies

and a separate plate, where the microcontroller could be mounted.

During the redesign, all parts were fused together and then were 3D-printed at the local

company, which vacuum-formed the front casing. Apart from mounting-holes for the

screws, which were adopted from the aluminium cases, cable guides, pockets for the

VOC-sensor and a possible sound level meter (not integrated) and louvers were

implemented.

4.3.3 Raspberry Pi Casing

We also decided to encase the microcontroller (see chapter 4.4) for the scent module in

a 3D-printed box (see Figure 49). The three reasons for this were:

• Protection (ESD problematic or purely mechanical damages)

• Preventing mishandling the 2nd USB-connector

• Easier fitting to the rear casing: instead of 4 screws, 4 nuts, 4 plastic spacers and 12

isolation rings, only the Raspberry PI casing, 2 screws and 2 nuts are required.

Mounting all pieces is now much easier, as no parts can come loose at any time

(the isolation rings and spacers tended to slip away when a screw was removed).


40

Figure 49: Microcontroller; left unprotected; right covered

4.3.4 Droplet shield

Before reworking the piston-mechanism, we reduced the amount of visible droplets by

creating removable shields. They were clipped onto the front casing and prevented

droplets from leaving the scent module and therefore also from creating stains on

furniture. The first design (reminiscent of a smiley-face) successfully caught most of the

droplets in a small integrated basin but was too airtight for adequate vaporization of the

scents.

The second design (which included bigger holes, projected from the front casing) was

designed to allow improved airflow past the caught scent, but the vaporization still was

not enough. The shields can be seen in Figure 47.

4.4 Electrical Design

The scent module is powered by a USB power supply rated at 5V/1A. The main electrical

components of the scent module are a controller board, a motor driver unit for the motors

and an analog digital converter (ADC) for system feedback. The actuators of the scent

module are controlled by a Raspberry PI Zero W board (see Figure 50), that allows for

connection to a building automation system over WLAN and handles the logics of

dispensing.

Figure 50: Raspberry Pi Zero W board computer


41

The motors of the scent module are connected to the Raspberry Pi Zero board via an

Adafruit DRV8833 motor driver board that features current limiting (both, for protecting the

power supply against overload in case the motor is locked, and against injuries) as well as

reverse voltage protection for the controller board. One DRV8833 board allows for the

connection of up to two motors (see Figure 51). The DRV8833 board is controlled over

digital outputs of the GPIO header of the Raspberry board and provides fault input via a

digital input.

a

Figure 51: Adafruit DRV8833 motor driver breakout board.

An Adafruit ADS1015 ADC module (featuring 12-bit resolution) is used to measure the

current flowing to the motors via a shunt resistor (see Figure 52). The ADC module is

connected via an I2C interface to the main controller board.

Figure 52: Adafruit 12bit ADC converter ADS1015 breakout board

By measuring the current flowing to the motors, the software can detect peaks and

values, which relate to the inverse current position of the pump spray. This allows for

automatic turn off at the upper position, which allows for easy removal and insertion of

scent bottles. Also, the exact number of actuations of the pump spray can be detected in

this way (see Figure 53 for typical current flow of the prototype loaded with bottles and

without bottles).


42

Figure 53: Current flow to motor during pump cycles with no bottle inserted (top), and bottle

inserted bottom) as raw signal and smoothed signal.

Figure 54 shows the detailed wiring of the DRV8833 and ADS1015 modules to the

Raspberry board.

Figure 54: Schematics for the electrical components of the scent module.

While the first prototypes built for functional testing were wired manually, a printed circuit

board (PCB) has been designed to make the assembly easier for the field test prototypes.

Figure 55 shows the circuit board layout, providing headers to directly plug the individual

boards together.


43

Figure 55: Printed circuit board design for the scent module electronics.

4.4.1 Environmental sensor

For the final field test design, we decided to put an environmental sensor into each scent

module, since it became clear, that variations in test settings required continuous

monitoring of air quality. The respective data is used to measure the effect dispensing

scent and to prevent oversaturation.

The environmental sensor module is mounted on the left top corner of the scent module

housing, pointing to the side, as seen in Figure 56 and in Figure 49. This position was

chosen, to prevent actual droplets from hitting the sensor and actually measure the

contents of the air without direct exposure to the spray channel.

Figure 56: VOC sensor mounting


44

Table 4 shows the hardware component list for the electrical parts of the scent module.

Table 4: GREAT Scent Module hardware component list


1 Adafruit DRV8833 motor driver

breakout board

1 Adafruit ADS1015 ADC breakout

board



(industrial)

2 Resistors 1Ohm ¼ Watt

2 Pololu Micro Metal Gear Motor 6V 30:1

1 Watterott Bosch BME680 breakout

board

4.5 Functional Tests

Concerning electronics, the functional and field tests were successful for the scent

module, no malfunctions, errors or weaknesses occurred.

5 Repeater

During adaptations to the field test setups at the testing places in Hall and Neumarkt,

reception issues of EnOcean telegrams between the controller and the luminaires have

been discovered. Due to the metal casing of the luminaire, and the unavoidable

electromagnetic noise produced by LED drivers, the EnOcean performance is significantly

reduced compared to e.g. PIR sensors which only have a plastic housing.

To work around this limited range of EnOcean telegrams, a special purpose repeater was

developed, that besides the standard EnOcean repeating functionality also offered a

logging functionality of received telegrams and their signal strength, to assist in finding the

best placements of the repeaters and reconstruct what was going on, if a luminaire did

not handle a switching command.


45

The GREAT Repeater is based on a Raspberry PI Zero with an EnOcean TCM310 transceiver

module (see Figure 57). The transceiver board is directly attached to the GPIO Header of

the Raspberry Pi Zero. The transceiver module can be configured to run as a level 1 (only

one step forwarding) or a level 2 (multistep forwarding) repeater and also allows for

setting filters (e.g. depending on the signal level) for which telegrams should be repeated.

This way EnOcean traffic caused by repeating signals can be reduced to a minimum. The

Raspberry Pi and the EnOcean module are placed inside a 3D printed housing (see Figure

58).

The software running on the Raspberry Pi reads received telegrams from the transceiver

module over the UART port and forwards them via a WLAN connection to a configurable

receiver.

Figure 57: EnOcean Repeater based on Raspberry Pi Zero and TCM310


46

Figure 58: 3D printed housing for the repeater

Table 5 shows the component list required for the GREAT Repeater.

Table 5: GREAT Repeater component list


1 EnOcean Pi Module with TCM310

1 3D Printed Housing



(industrial)

6 Sensor Technology

6.1 Sensors for acquiring activation and relaxation status

The sensors used to detect changes in the activation and relaxation level of the group

follow two approaches. Ambient PIR (passive infrared) sensors measure group activity

based on frequency of motion detection. As soon as at least one person is in the room,

the system uses the PIR sensors to continuously determine the relative activity level in the

room (estimated number of entries/exits per time unit, estimated number of persons, total

movement per area and person). The PIR data cannot be assigned to any person.

Body worn sensors are in use for caregivers to measure activity level and stress level.

Wearables and smart textile for the patients are not in use due to the vulnerability of the

target group and expected difficulties with acceptance and adherence.


47

6.1.1 PIR Sensor

The standard use case for PIR sensors is triggering lighting on basis of detected movement

or presence. Part of further investigation is the detection of inactivity. Can the presence

of inactive presence be recognized by history knowledge about the number of persons

entered the room? For GREAT the main aspect is monitoring of activity and detection of

group activity levels.

For the activation of lighting the sensor is typically triggered once by motion or presence

and switches the light. For a period of 60-90 seconds or more the light is activated and the

sensor falls into a sleep mode to conserve energy. After that sleep period the sensor starts

to detect activity again. If no activity is detected, the light is switched off otherwise the

light stays on and another sleep period starts.

Figure 59: Detection range

The requirement for GREAT is to continuously detect activity. Therefore, the sleep period

has to be as short as possible. To support convenient and modular installation the second

requirement is wireless technology. Since there is a conflict between a wireless and

energy-saving architecture and high frequency detection the chosen product has to be

modified. The minimum sleep period was reduced to about 1 second. An additional

modification is made due to the fact, that the sensor is placed on top of the ceiling and

not as concealed installation (see Figure 60). If the human heat source moves through the

individual zones, different charge differences are generated on the associated sensor

elements and a movement can be detected over a large. The main issue would be to

recognize not only movements and classify a single person’s activities as shown in (Nef et

al. 2015; Luo et al. 2017), but also how group activities can be classified and what this

means related to a stressful or too calm situation. The sensor uses an optical shell

combined with a magnetic cap to provide easy access in case of battery change

without the necessity to completely removing the sensor. For GREAT the Thermokon

“EasySens” SR-MDS BAT will be used (for technical specifications see Table 6, picture see

Figure 60).


48

Table 6: Technical specification for Thermokon SR-MDS BAT

Vendor Thermokon Sensortechnik GmbH

(Germany)

Series EasySense

Type SR-MDS BAT

Retail price 250,00 EUR

Technical design Wireless

Wireless technology EnOcean ISO/IEC 14543-3-10

Radio frequency 868,3 MHz

Functions Motion detection and brightness

measurement

Motion detection Passive infrared

Detection area 360°; 105° conical (ceiling installation)

Detection radius (2,5 m room height) 3,25 m

Power source 3 x Battery 3,6V 1/2 AA LS14250

Brightness (Accuracy) 0-510 Lux (+/- 30 Lux)

Sleep time interval (modified) 1s – 1000s


49

Figure 60: Thermokon SR-MDS BAT with extra light 3D printed cover

6.1.2 Body worn sensor

Changes in activity and stress or relaxation level can be measured by body worn sensors,

which measure corresponding physiological parameters like heart rate, skin conductance

or motion patterns via accelerometers. The Everion sensor from the company Biovotion

(Zurich, Switzerland) makes use of various sensing techniques which can all be applied on

the upper arm. There are several optical channels based on different colours to detect

changes in subcutaneous tissue and a galvanic skin response sensor (GSR). Based on this

raw data the human understandable vital parameters are calculated. The vitals available

are listed in Table 7. The most interesting and most often used parameter for stress

recognition is the heart rate variability (HRV). It is calculated based on the beat to beat (R-

R interval) as illustrated in Figure 63. The HRV of a well-conditioned heart is typically large

at rest. It might decrease in case of activity or interesting for us in case of mental stress.

Biovotion applies to root mean squared of successive differences (RMSSD) to calculate

the HRV. For our algorithm to recognize stress phases as shown in Figure 63, we apply three

functions on the R-R intervals (according to Ulrich Reimer et al., 2017):

• SDNN: standard deviation of RR intervals (i.e., intervals between two heart beats)

• RMSSD: root mean square difference of successive RR intervals in the time frame

• PNN50: percentage of pairs of adjacent RR intervals differing by more than 50 ms.


50

Figure 61: Biovotion Everion upper arm

sensor

Figure 62: Everion back with light and

galvanic skin sensors

Figure 63: HRV signal and marked high stress segments

Therefore, we need the R-R values in milliseconds which are provided every second (1Hz).

Together with the main vitals, a quality value in the range of 0-100 is provided. Values with

a quality below 50 are ignored.

The algorithm applied in detail to derive the stress segments with the result shown in Figure

63 is described in detail in (Ulrich Reimer et al. 2017).

Another interesting testing parameter for our application might be the GSR, also referred

to as Electrodermal Activity (EDA), which measures changes when starting to sweat. This is

commonly used as a sensitive measure for emotional arousal. The GSR provided every

second is expressed in a value between 0 – 65535 and has a conversion factor of 1/3000.

The advantage of the Everion in comparison to other sensors is its approval in Europe as a

medical device for heart rate (HR) and blood oxygenation or oxygen saturation (SpO2)

and its ISO certifications (ISO 1345). An approval by the FDA in the USA is currently

ongoing. Also, the higher level of acceptance is an advantage which has been shown in

HR

HRV


51

our tests at the sleep laboratory where the devise has been applied by over 40 healthy

test persons and over 30 unhealthy persons in parallel to ECG with electrodes attached

and wrist worn devices (Reimer et al. 2017). The sensor is lightweight and convenient to

wear, can store data locally, transmit data in real-time or whenever in range of a

gateway and can be recharged easily by placing the sensor onto a conductive charging

cradle. The comfortable wearing is supported by an elastic textile band available in

different sizes as shown in Figure 64.

Figure 64: Different sizes of bands for an optimal fit

Table 7: Technical specification for Biovotion Everion

Vendor Biovotion AG (Switzerland)

Type Everion

Price (project) 550,00 EUR

Technical design Wireless with rechargeable battery

Wireless technology Bluetooth 4.0 + LE (IEEE 802.15.1)

Transmission range <10 m

Radio frequency 2,4 GHz

Parameters Heart rate

Blood oxygenation

Skin temperature

Skin blood perfusion

Steps / Motion


52

Experimental parameters (project) Respiratory rate

Heart rate variability

Energy expenditure

Blood pulse wave

Skin conductance

Data modes Vital sign parameters, raw data,

mixed mode

Battery life 24 h

Power source Embedded Li-Ion battery

rechargeable

The sensor can transmit pre-calculated vital sign data or raw data from every parameter

channel in real-time or buffer it to the internal memory. The memory can hold several days

of vital sign data or 4 hours of raw data. The sampling rate of raw data at 53 Hz with 12

different channels causes a heavy data volume which leads to asynchronous data

transmission over time (transmission is slower than recording of raw data).

In GREAT the vital sign mode is the preferred mode since the basic vital parameters are

calculated in real-time for detection of changes in activation/relaxation and physical

activity levels. For receiving data, a gateway has to be installed. Different variations are

currently evaluated and considered for use. The preferred gateway is a Raspberry Pi 3

board controller which is also in use as the main controller of the GREAT system.

Alternatives can be smartphones with an Android operating system or a Raspberry Pi Zero,

which is the cheapest of all variations. The vendor currently only supports Windows and

Android operating systems. Those variations rely on more expensive hardware like Stick-

PCs based on Intel architecture, which is in conflict with a cheap end user price. To

provide maximum compatibility and avoid problems with different Bluetooth stacks of

different hardware components the vendor developed his own BT-Dongle. This dongle is

also expensive (120 CHF). Different alternatives including advantages and disadvantages

as well as the price tag are listed below (Figure 65).

For the functional tests, we started with the PC-Stick version and transfer the collected

vitals every evening to our Cloud Server where also the logs from all devices are stored.

Originally, for the field tests, we planned for the cheaper variant with a Raspberry Pi 3

running Android. However, while porting the software, unexpected driver issues occurred.

Therefore, we kept the working solution from the functional tests for the field test

prototypes.


53

Variation Pro/Cons Price

(CHF)

Intel Stick-PC with Windows 10 and BT-Stick

Pro:

Use of existing

software

Cons:

Expensive

Stability issues

250

Raspberry Pi 3 with Windows

Pro:

Cheap

Cons:

Porting of

software

170

Raspberry Pi Zero with onboard BT

Pro:

Cheap

Board in use in

other modules

Cons:

Porting of

software

30


54

Raspberry Pi 3 with Android

Pro:

Cheap

Software under

development by

vendor

Cons:

Stability issues

60

Raspberry Pi 3 with Wine Emulation

Figure 65: Possible hardware combinations incl. price tags

6.2 Environmental Sensor

In order to get more information about the environment where the GREAT system is being

used and integrated, environmental sensor unit (BOSCH BME 680) has been incorporated

into the scent module using a ready-made breakout board for easier placement and

connection (see Figure 66).

The environmental sensor unit is connected to the controller of the Scent module via an

I2C interface. It allows for measurements of air quality (VOC - Volatile Organic

Compounds), temperature, pressure and relative humidity (see Table 8).

Figure 66: Watterott Bosch BME 680 breakout board


55

Table 8: Bosch BME 680 Environmental Units Specification

Interface I²C and SPI

Gas sensor

Response time (τ

33-63%)

Sensor-to-sensor

deviation

Power

consumption

Output data

processing

< 1 s (for new sensors)

+/- 15% +/- 15

< 0.1 mA in ultra-low power mode

direct output of IAQ: Index for Air Quality

Humidity sensor

Response time

(τ0-63%)

Accuracy

tolerance

Hysteresis

8 s

± 3 % relative humidity

≤ 1.5 % relative humidity

Pressure sensor

RMS Noise

Sensitivity Error

Temperature

coefficient offset

0.12 Pa (equiv. to 1.7 cm)

± 0.25 % (equiv. to 1 m at 400 m height change)

±1.3 Pa/K (equiv. to ±10.9 cm at 1°C temperature change)


56

7 Controller

The main controller is based on a Raspberry Pi 3 single board-computer (Figure 67). It

features a Broadcom BCM2837 system on a chip with four ARM Cortex-A53 cores clocked

at 1.2 GHz. It includes 1 GB of LPDDR2 RAM and features a 10/100 Ethernet port, a 2.4 GHz

802.11n wireless module and a Bluetooth 4.1 LE module. For extension, it offers 4 USB2 ports

and a 40-pin GPIO header (e.g. providing support for I2C, I2S, SPI, UART interfaces).

To integrate EnOcean components, a Raspberry Pi EnOcean module based on the

TCM310 EnOcean transceiver module is connected to the built in UART via the GPIO

header of the Raspberry. For better signal performance of the EnOcean modules, an

external 868 MHz antenna was originally connected to the EnOcean module, however,

tests during the field trials have shown that the integrated antenna of the EnOcean

module would be sufficient, if it’s straightened. Table 9 lists the hardware components

required for the GREAT controller.

Figure 67: Controller setup based on Raspberry PI 3 and EnOcean PI 868 MHz module with external

antenna on the left, and wire-antenna on the right.


57

Table 9: GREAT Controller hardware component list

1 Raspberry Pi 3 board

1 Casing for Raspberry Pi 3

1 Micro USB Power Supply, 2.5A

1 microSDHC card, 16GB, Class 10 (industrial)

1 EnOcean PI 868 module for Raspberry

1 868MHz Antenna H-Tronic 1618110

1 SMA-Connector Socket

The reasons for choosing the Raspberry Pi 3 board as basis for the controller instead of

other single board computers (e.g. like the Beagle Bone boards) were price, strong

community support including a wide range of extensions, and most importantly built-in

wireless support.

For a documentation of the software architecture and features of the field test prototype

see Deliverable 2.4.


58

8 References

[1] Lighting concept based on literature search see D1.1

[2] Carvalho, DZ., St Louis, EK., Knopman, DS., Boeve, BF., Lowe, VJ., Roberts, RO.,

Mielke, MM., Przybelski, SA., Machulda, MM., Petersen, RC., Jack, CR., Vemuri, P.

(2018) Association of excessive daytime sleepiness with longitudinal β-

Amyloid Accumulation in Elderly Persons Without Dementia. JAMA Neurol. 75(6):

672-680.

[3] Shokri-Kojori. E., Wang, GJ., Wiers, CE., Demiral, SB., Guo, M., Kim, SW., Lindgren, E.,

Ramirez, V., Zehra, A., Freeman, C., Miller, G., Manza, P., Srivastava, T., De Santi, S.,

Tomasi, D., Benveniste. H., Volkow, ND. (2018) β-Amyloid accumulation in the

human brain after one night of sleep deprivation. Proc Natl Acad Sci USA 115(17):

4483-4488. doi: 10.1073/pnas.1721694115.

[4] Ulrich Reimer; Emanuele Laurenzi; Edith Maier; Tom Ulmer (2017): Mobile Stress

Recognition and Relaxation Support with SmartCoping: User-Adaptive

Interpretation of Physiological Stress Parameters Hilton Waikoloa Village, Hawaii,

USA, January 4-7, 2017. In: 50th Hawaii International Conference on System

Sciences, HICSS 2017, Hilton Waikoloa Village, Hawaii, USA, January 4-7, 2017: AIS

Electronic Library (AISeL). Online verfügbar unter http://aisel.aisnet.org/hicss-

50/hc/apps_for_health_management/5.

[5] Reimer, Ulrich; Emmenegger, Sandro; Maier, Edith; Zhang, Zhongxing; Khatami,

Ramin (2017): Recognizing Sleep Stages with Wearable Sensors in Everyday Settings.

In: Proceedings of the 3rd International Conference on Information and

Communication Technologies for Ageing Well and e-Health. 3rd International

Conference on Information and Communication Technologies for Ageing Well and

e-Health. Porto, Portugal: SCITEPRESS - Science and Technology Publications, S. 172–

179.


59

9 List of figures

Figure 1: GREAT Components Overview, Source: GREAT consortium ...................................... 6

Figure 2: Schematic representation of the lighting solution in GREAT. The biodynamic curve

is running automatically, light interventions (activation, relaxation, TV scene) are

adjustable manually or automatically ........................................................................................ 8

Figure 3: Biodynamic lighting concept. CCT…colour temperature in Kelvin, Eh…horizontal

illuminances of the task light, EhRoom…horizontal illuminance of ambient room light

(resulting from Ev + Eh Task, for clarity reasons EhRoom is not shown), Ev…vertical illuminance

at the eye level ............................................................................................................................. 9

Figure 4: GREAT lighting concept ................................................................................................ 9

Figure 5: Activating light cue ..................................................................................................... 10

Figure 6: Relaxing light cue ........................................................................................................ 11

Figure 7: Different control options for the GREAT luminaire: basic functions (left), advanced

functions (right) ........................................................................................................................... 11

Figure 8: Increasing the vertical illuminance by adding a diffuse panel to the luminaire

(left) and reaching high horizontal illuminances by adding task lighting by spots (right) .... 12

Figure 9: Results of light pressure study...................................................................................... 13

Figure 10: Optical concept for the GREAT Uplight (left) and the 2 possible Luminous

Intensity Distribution curves LID (right): The Indirect light is available as an asymmetric

solution for the wall mounted luminaire and as a symmetrical light distribution for example

for the free-standing luminaire situated in the middle of the room ....................................... 14

Figure 11: Picture of the Uplight of the GREAT luminaire ......................................................... 14

Figure 12: The LEDs are placed on the edge for lateral input of light in the light guiding

plate ............................................................................................................................................ 15

Figure 13: Dialux simulation of the GREAT Uplight .................................................................... 15

Figure 14: Downlight of the GREAT luminaire and LPG ............................................................ 16

Figure 15: Dialux simulation of the GREAT Uplight and Downlight .......................................... 16

Figure 16: Spot of the GREAT luminaire ..................................................................................... 17

Figure 17: Dialux simulation of the GREAT Uplight, Downlight and Spot ................................ 17

Figure 18: Dialux simulation of the GREAT Uplight, Downlight and Spot, results The values

listed in the table are already achieved with warm white LEDs (2200K). With cold white

LEDs (5000K) higher illuminances can be achieved. ............................................................... 18

Figure 19: Technical specification sheet GREAT luminaire ...................................................... 19

Figure 20: Dialux simulation of the GREAT luminaire showing the illuminances in a 25 m²

room. ........................................................................................................................................... 20


60

Figure 21: Dialux simulation of the light distribution of the GREAT luminaire. Also shown are

the isolines for 150 lx (orange), 300 lx (yellow) and 600 lx (grey) horizontal illuminance. ..... 20

Figure 22: Dialux simulation of the GREAT luminaire showing the luminances in a 25 m²

room. ........................................................................................................................................... 21

Figure 23: Free-standing, mobile GREAT luminaire ................................................................... 21

Figure 24: Wall-mounted GREAT luminaire in a patient's room in the hospital in Hall (Austria)

..................................................................................................................................................... 22

Figure 25 : The buckling mechanism of the free-standing luminaire ...................................... 22

Figure 26: Special adjustment for psychiatric institutions: breaking point for protection

against suicide ............................................................................................................................ 23

Figure 27: GREAT system implemented in a patient's room in the hospital in Hall (Austria) . 24

Figure 28: GREAT standing luminaire in the common area in the hospital in Hall (Austria) .. 24

Figure 29: New, optimized luminaire head with EnOcean board and driving board .......... 25

Figure 30: EnOcean switch ........................................................................................................ 26

Figure 31: EMV measurement of the GREAT luminaire for CE conformity .............................. 26

Figure 32: Transportable GREAT Luminaire ............................................................................... 27

Figure 33: Easy adjustment of GREAT luminaire ....................................................................... 28

Figure 34: Cable storage feature .............................................................................................. 28

Figure 35: GREAT sound module based on Logitech Z150 active speaker ........................... 29

Figure 36: Raspberry PI and Pi-DAC Zero integrated into the Z150 active speaker. ............. 30

Figure 37: 3D model of Raspberry PI + Pi-DAC Zero holder for mounting inside Z150 speaker

..................................................................................................................................................... 31

Figure 38: Wiring of the Pi-DAC 2x4 header to the Logitech board inside the speaker ....... 31

Figure 39: 3D model of separate housing for sound module logics ....................................... 32

Figure 40: External sound module with a Pioneer RM05 Speaker featuring a frequency

range of 40-50.000Hz .................................................................................................................. 33

Figure 41: Speaker covers to prevent accidental adjustment of volume level .................... 33

Figure 42: Sine, linearization and database for linearization and formula for the linear

approximation ............................................................................................................................ 35

Figure 43: Error between sine and linear approximation, absolute [-1, 1] and percentage 35

Figure 44: Evolution of mount (top), piston (center left), eccentric (center right) and lever

(bottom) ...................................................................................................................................... 36

Figure 45: Left shows the no-load-phase of the motor, right shows the spring-tensioning-

phase of the motor ..................................................................................................................... 37

Figure 46: Reworked mechanic installed in the rear casing ................................................... 37


61

Figure 47: Evolution of the 3D-printed front casing (top) and the two variants of shields

(bottom) ...................................................................................................................................... 38

Figure 48: Left shows the completely mounted rear casing, right shows the front casing ... 39

Figure 49: Microcontroller; left unprotected; right covered .................................................... 40

Figure 50: Raspberry Pi Zero W board computer ..................................................................... 40

Figure 51: Adafruit DRV8833 motor driver breakout board. .................................................... 41

Figure 52: Adafruit 12bit ADC converter ADS1015 breakout board ....................................... 41

Figure 53: Current flow to motor during pump cycles with no bottle inserted (top), and

bottle inserted bottom) as raw signal and smoothed signal. ................................................. 42

Figure 54: Schematics for the electrical components of the scent module. ........................ 42

Figure 55: Printed circuit board design for the scent module electronics. ............................ 43

Figure 56: VOC sensor mounting ............................................................................................... 43

Figure 57: EnOcean Repeater based on Raspberry Pi Zero and TCM310 ............................. 45

Figure 58: 3D printed housing for the repeater ........................................................................ 46

Figure 59: Detection range ........................................................................................................ 47

Figure 60: Thermokon SR-MDS BAT with extra light 3D printed cover ..................................... 49

Figure 61: Biovotion Everion upper arm sensor ......................................................................... 50

Figure 62: Everion back with light and galvanic skin sensors .................................................. 50

Figure 63: HRV signal and marked high stress segments ......................................................... 50

Figure 64: Different sizes of bands for an optimal fit ................................................................ 51

Figure 65: Possible hardware combinations incl. price tags ................................................... 54

Figure 66: Watterott Bosch BME 680 breakout board .............................................................. 54

Figure 67: Controller setup based on Raspberry PI 3 and EnOcean PI 868 MHz module with

external antenna on the left, and wire-antenna on the right. ............................................... 56


62

10 List of tables

Table 1: Advantages and disadvantages of diffuse and task lighting .................................. 12

Table 2: GREAT Light Module hardware component list ......................................................... 28

Table 3: GREAT Sound Module hardware component list ...................................................... 32

Table 4: GREAT Scent Module hardware component list ....................................................... 44

Table 5: GREAT Repeater component list ................................................................................. 46

Table 6: Technical specification for Thermokon SR-MDS BAT .................................................. 48

Table 7: Technical specification for Biovotion Everion ............................................................ 51

Table 8: Bosch BME 680 Environmental Units Specification ..................................................... 55

Table 9: GREAT Controller hardware component list .............................................................. 57

great – aal-2016-023 · great – aal-2016-023 34 4 scent module compared to the field trial...

Documents